Using Ontology to Harmonize Knowledge Concepts in
Data and Models
Wien, J.J.F. 1, M.J.R. Knapen 1, S.J.C. Janssen 2, P.J.F.M. Verweij 1, I.N. Athanasiadis 3, H. Li 3, A.E.
Rizzoli 3 and F. Villa 4
1 Alterra, Environmental Sciences Group, Wageningen University and Research Centre, Wageningen
2 Wageningen university, Plant Production Systems Group, Wageningen
3 Dalle Molle Institute for Artificial Intelligence (IDSIA), Lugano
4 University of Vermont, Vermont
Email: jan-erik.wien@wur.nl
Keywords: ontology, modelling, model integration, integrated assessment
EXTENDED ABSTRACT
Policy makers are confronted with complex,
socially relevant problems. To increase the quality
of their policy and realize a broader support for
and understanding of actions, the policy making
process has become a participatory process where
policy measures need to be assessed in an
integrated context (Rotmans and van Asselt
,1996).
In these integrated assessment studies, people
work from different perspectives and domains, e.g.
from an agricultural modelling perspective, an
environmental perspective or an economic
policy/problem perspective. Semantic
interoperability is the key factor to integrate the
knowledge of these different domains and
perspectives in a computerized framework.
Semantic interoperability is the ability of
systems/components to share and understand
information at the level of formally defined and
mutually accepted domain concepts (Sølvberg,
1998). The specification of these concepts is done
with ontology.
One of the most cited definitions of ontology is
from Gruber (1993): “an explicit and formal
specification of a conceptualization”. A
“conceptualization” can be considered as an
abstract model for a phenomenon identifying the
relevant concepts. All concepts are explicitly
described in a formal machine readable language.
The challenge in integrated modeling is the
conceptual integration. To achieve this, we need
explicit semantics and a shared conceptualization.
For this we need to tackle the different perceptions
and interpretations of people involved. Different
modeling approaches, different formalism and last
but certainly not least, the different integration
requirements and ambitions need to be taken into
account.
In the SEAMLESS project (Van Ittersum et al.,
2007), the process used for creating a common
ontology for models, indicators and raw data is
based on a participatory and collaborative
approach. A dedicated taskforce was created with
participants from different parts of the project, and
coordinated by the work package charged with
integration. This task force envisages to develop a
common knowledge base that represents a shared
conceptualization between the different databases,
models and indicators, with adequate meta-data.
A core component, called the SEAMLESS
Knowledge Manager (KM) (Villa et al., 2007),
provides functionality as an extensible semantic
modeling toolkit. It loads meaning through OWL
(Ontology Web Language) ontology and
transparently connects formal concepts to software
objects and literals. Scalable use of machine
reasoning effectively integrates an object-oriented
framework, an object oriented database system and
an ontology-based knowledge management
environment into a SEAMLESS whole.
The development of the SEAMLESS common
ontology was and still is a big challenge. By
putting ontology in a central position in the project
and the systems architecture, this shared
conceptualization is the basis for generating (Java)
source code for the object classes representing all
the concepts and representing the objects in
relational database tables. The use of ontology has
proved to be very useful if not essential both for
the technical integration of knowledge in the
SEAMLESS Integrated Framework and in
understanding the meaning of communicated
words of the diversity of people within the project.

1959

1. INTRODUCTION
Policy makers are confronted with complex,
socially relevant problems. By handling these
problems, it is a general tendency to increase
participation of citizens and other stakeholders. In
this way policy makers want to increase the quality
of their policy and realize a broader support for
and understanding of actions. In this new
governance concept, the policy making process is
the product of complex interactions between
governmental and non-governmental
organizations, each seeking to influence the
collectively binding decisions that have
consequences for their interest. Policy making is
more and more a process of cooperation and
participation in which the policy maker becomes a
facilitator of the process. This concept is based on
the assumption of the model of “co-production of
knowledge” (Callon, 1999).
To account for this new governance, policy
measures need to be assessed in an integrated
context. Rotmans and van Asselt (1996) defined
Integrated Assessment as “an interdisciplinary and
participatory process combining, interpreting and
communicating knowledge from diverse scientific
disciplines to allow a better understanding of a
complex phenomena. In this interdisciplinary and
participatory process, information needs to be
accessible in the way that all different types of
stakeholders achieve a mutual understanding of the
problems, objectives and solutions. But this mutual
understanding across disciplines is often hindered
by jargon, language, past experiences and
presumptions of what constitutes persuasive
argument, and different outlooks across disciplines
or experts of what makes knowledge or
information salient for policy makers or policy
assessments (Cash et al., 2003).
This paper will describe the problems and (partial)
solution of conceptual misunderstanding and
knowledge integration in Integrated Assessments
both from a theoretical perspective (section 2) and
practical experiences with ontology based model
integration (section 3). It will explain the
challenge and the process used in a number of
European integrated projects (e.g. SEAMLESS,
AquaStress), for creating an ontology for (linking
of) projects, models, indicators and raw data.
Section 4 gives a description of the role of
ontology in the architecture and design of the
integrated framework of SEAMLESS.

2. CHALLENGE IN INTEGRATING
KNOWLEDGE
2.1. Theoretical perspective
Interoperability is the ability of two or more
systems or components to exchange information
and to use the information that has been exchanged
(IEE,1990). There is often a distinction between
syntactic, structural and semantic interoperability
(Sølvberg, 1998).
• Syntactic interoperability is defined as the ability
of two or more systems/components to exchange
and share information by marking up data in a
similar fashion (e.g. using XML).
• Structural interoperability means that the
systems/components share semantic schemas (data
models) that enable them to exchange and
structure information (e.g. using RDF).
• Semantic interoperability is the ability of
systems/components to share and understand
information at the level of formally defined and
mutually accepted domain concepts.
Semantic interoperability requires the correct
interpretation and mutual understanding of all
transferred information. In order to obtain mutual
understanding of interchanged data, the actors
have to share a model of what the data stand for.
Semantic interoperability is about how to achieve
such mutual understanding (Sølvberg, 1998).
One of the earliest theories dealing with
understanding and remedies for misunderstanding
is Richards's Meaning of Meaning Theory (Ogden
and Richards, 1923). Instead of focusing on the
information that is communicated, Richards
wanted to study the meaning of the words. He felt
that understanding is the main goal of
communication and communication problems
result from misunderstanding.
One of the ideas behind the Meaning of Meaning
Theory is "The Proper Meaning Superstition"
(Ogden and Richards, 1923). This is the false
belief that every word has an exact, "correct"
meaning. Richards says that the Proper Meaning
Superstition is false because words mean different
things to different people in different situations.
This misunderstanding causes problems when two
people believe they are talking about the same
thing, but actually talk about different things.
Another concept that Richards uses is the idea of
signs and symbols in communication. Words are
examples of such symbols. Symbols have no
1960

natural connection with the things they describe
(Griffen, 1997). Words are symbols of something
because they have been given meaning. But very
often words mean one thing in a certain context
and mean another thing in a different context. This
is why it is so important to study the context to get
a better understanding of the meaning.
To come to this better understanding, Richards
invented the Semantic Triangle. This triangle
shows the relationship between symbols and their
referent. One part of the triangle is the symbol, or
the word. Another peak on the triangle is the
thought or reference. This is the words that one
would use to describe the referent. The referent,
the last part, is the thing that one would picture in
his mind.

Figure 1. Semantic Triangle by Richards (Ogden
and Richards, 1923).

An example in integrated agricultural modeling
would be “wheat”. The words to describe the
referent can be for example “cereal crop, grain,
flour”. The referent, the last part, is the thing that
one would picture in his mind.

Figure 2. Semantic Triangle for wheat.


Understanding that people mean different things
when they say the same thing is an important
concept for people to understand. Richards gives
ways to solve this problem of ambiguity. One of
them is to give a definition. Definitions are words
used in place of another word to explain the
thought in a person's mind. Another option to
understand the meaning by using a metaphor. A
metaphor can help to clarify what each person is
saying.
Feed forward is also an important factor when
trying to avoid misunderstanding. Feed forward is
when the speaker thinks of how his audience will
react to what he is about to say and adjusts his
words accordingly (Ogden and Richards, 1923).
Feed forward forces the speaker to consider the
experiences of the audience in order to better
explain what they are saying.
To enable semantic interoperability in integrated
modeling, the problem of semantic conflicts or
semantic heterogeneity needs to be solved. For this
ontology will be used. The term ontology is
borrowed from philosophy, where ontology is a
systematic account of Existence.
One of the many definitions of ontology is from
Neches et al. (1991) “an ontology defines the
basic terms and relations comprising the
vocabulary of a topic area as well as the rules for
combining terms and relations to define extensions
to the vocabulary”. One of the most cited
definition is from Gruber (1993): “an explicit and
formal specification of a conceptualization”.
A “conceptualization” is explained as an abstract
model of a phenomenon, identifying the relevant
attributes. A conceptualization is an abstract,
simplified view of the world that we wish to
represent. Every knowledge base or knowledge-
based system is committed to some
conceptualization. (Gruber 1995)
“Formal specification” refers to the fact that the
language semantics are machine readable. Often
this is done by use of W3C OWL (Ontology Web
Language, Patel-Schneider et al., 2004). A formal
specification helps to communicate the definition
of terms in a context independent ways and formal
language semantics allows some automated
consistency checks.

2.2. Practical experiences of ontology based
knowledge integration
In integrated assessment projects a large number of
scientists co-operate working on a wide range of
issues, like different type of models, data and
1961

databases, indicators, assessment problems and
user interaction. This wide range of activities
needs to be brought together, as the models should
be using the data, the indicators should be based
on model outputs and data, while the assessment
problem needs to be consistently defined to
parameterize or configure the models coherently.
Also indicators need to link to the assessment
problem to be clearly presented to the end-users.
This leads to a complex integration problem in
which many scientists from different domains
should contribute to achieve one shared
understanding of the integrated assessment
procedure.
In such a complex integration task, different types
of misunderstandings around the meaning of
concepts can occur:
1. as the same concepts might be used
for different meanings, for example
area in a model and area in the
database,
2. as different concepts might be used,
which have the same meaning, for
example an internal user and an
integrative modeler,
3. as concepts might be used with an
ambiguous meaning, for example
scenario (Shoemaker, 1993),
4. As relationships between concepts
might be understood in a different
way, for example between the
different spatial scales and
administrative regions.
The challenge in integrated modeling is the
conceptual integration. To achieve this, we need
explicit semantics and a shared conceptualization.
For this we need to tackle the different perceptions
and interpretations of people involved. Different
modeling approaches, different formalism and last
but certainly not least, the different integration
requirements and ambitions need to be taken into
account.
3. USE OF ONTOLOGY
3.1. Common Ontology
Ontology helps to formalize the knowledge
captured in and/or between models, in order to
subsequently facilitate model development, testing
and documentation (Scholten and Kassahun) and
model re-usability and exchangeability (Rizzoli et
al., 2005) and separates knowledge captured in the
model from the actual implementation in a
modelling language or software e.g. Java,
FORTRAN, Mathlab, STATA, etc. (Gruber, 1993;
Villa et al., 2006) or from the data in a database
(Zander & Kächele, 1999).
The development of a common ontology by a
group of researchers is a complex, challenging and
time-consuming task (Farquhar et al, 1995;
Gruber, 1993), that still remains a scientific
challenge. Tools are available that help in ontology
development and store the ontology once it was
developed. To achieve ontological commitment,
i.e. the agreement by multiple parties to adhere to a
common ontology, when these parties do not have
the same experiences and theories (Holsapple &
Joshi, 2002) a collaborative approach should be
used. Other approaches for ontology development
are the inspirational approach, the inductive
approach, the deductive approach and the synthetic
approach (Holsapple & Joshi, 2002). A
collaborative approach has the advantages that
researchers from different disciplines are diverse in
their contributions, which avoids blind spots and
which has more chances of getting a wide
acceptance (Holsapple & Joshi, 2002) and that it
can incorporate the other approaches, e.g. synthetic
approach, as required for development of parts of
the ontology.
3.2. Ontology Engineering
SEAMLESS is an integrated assessment modelling
project (Van Ittersum et al., 2007), which aims to
provide a computerized framework to assess the
sustainability of agricultural systems in the
European Union at multiple scales. The process
used in the SEAMLESS project, for creating a
common ontology for models, indicators and raw
data is based on a participatory and collaborative
approach. A dedicated taskforce was created with
participants from different parts of the project, and
coordinated by the Work Package charged with
integration. This task force, called the DOT.force
(Data and Ontology Taskforce) envisages
developing a common knowledge base that
represents a shared conceptualization between the
different databases, models and indicators, with
adequate meta-data. The DOT.force has
knowledge engineers and domain members. These
knowledge engineers have knowledge and
experiences in the design and content of either
databases or ontology. Domain members will be
flexibly involved by this core group of knowledge
engineers to develop specific parts of the ontology
or database. The domain members hold knowledge
about a specific domain, like a model, a database,
indicators, scenarios, or the SEAMLESS-IF, which
should be captured by the knowledge base and the
database.
1962

The knowledge engineers in the DOT.force initiate
on a number of actions that should ultimately lead
to a complete ontology. These actions are:
1. integrating the different databases into
one SEAMLESS database,
2. clarifying the interfaces between the
models, while adding relevant meta data,
3. linking indicators to model outputs in the
ontology and retrieving required data
from the database,
4. supplementing the ontology with
additional meta-data on the concepts it
holds, like units, minimum and maximum
value, source and references,
5. developing a common ontology to cover
concepts relevant for runs with the
SEAMLESS-IF like scenarios, projects,
scales, problems, running time, etc
Different methods are used to construct the
ontology for the different actions. For actions 1
and 2 on databases and models, dedicated meetings
are organized to develop the ontology, while for
action 3 on indicators a proposal was made by the
knowledge engineers, which was then evaluated by
relevant domain members. Action 4 on metadata is
carried out independently by domain members,
once agreement on the common ontology has been
reached between domain members and knowledge
engineers. For action 5 on project and scenario
definition an iterative process was used to develop
a document, which is synchronized with a project
ontology after each iteration.
4. SYSTEM ARCHITECTURE
This chapter describes how ontology plays a
central role in the architecture of SEAMLESS-IF.
A core component, called the SEAMLESS
Knowledge Manager (KM), provides functionality
as an extensible semantic modeling toolkit. It loads
meaning through OWL ontology’s and
transparently connects formal concepts to software
objects and literals. Scalable use of machine
reasoning effectively integrates an object-oriented
framework, an object oriented database system,
and an ontology-based knowledge management
environment into a SEAMLESS whole. A tight
API supports storage, search, retrieval and
advanced management of semantically explicit
software objects.
The use of explicit semantics in software allows
attaining integration goals that are relevant to
many disciplines and applications. By means of
plug-in packages, the Knowledge Manager can be
extended with knowledge and software
functionalities to support specific semantic
modeling tasks. The plug-in packages include for
example:
1. Ontology and software support to handle
measurement units, datasets, observations and re-
definable spatial and temporal contexts;
2. Ontology and software support to handle several
representations of time and space, including
support for GIS data formats in both raster and
vector form;
3. The OPAL framework to facilitate specification
of semantic objects in automatically generated,
customizable XML schemata;
4. Support for repositories of semantic objects that
wrap common data formats and RDBMS/SQL
database front-ends.

Figure 3. Knowledge manager (Villa et al. 2007,
SEAMLESS PD 5.4.2.2).
The concepts relevant to the SEAMLESS domain
(mainly the agricultural domain) and the models
used in the SEAMLESS project have been put into
ontology. This ontology, together with related ones
(e.g. for measurement units), are loaded into the
Knowledge Manager component. The Knowledge
Manager can make the links between the models.
The actual exchange of information is based on the
Open Modeling Interface (OpenMI) (Gijsbers et
al., 2005; Gijsbers et al. 2006) that provides a
standardized interface to define, describe and
transfer data between software components that
run sequentially. This choice was made based on
1963

technical and functional requirements and the
possibility to re-use legacy models.
In the current second prototype version of the
SEAMLESS software the facilities of the
Knowledge Manager are used off-line (not in
runtime) to generate (Java) source code for the
object classes representing all the concepts, and
matching object-relational mapping files for use
with Hibernate (www.hibernate.org).
In the generated source code Java annotations are
used to record the connection to the ontology, like
in this example:
@ConceptURI("http://localhost/ontologies/c
rop.owl#Crop")
public class Crop implements Serializable
{
....
@PropertyURI("http://localhost/onto
logies/crop.owl#hasCropSoilRequirements")
Public CropSoilRequirements
getCropSoilRequirements()
....
}
Since the annotations are accessible at runtime the
ontology information can be used for reasoning,
e.g. to validate whether an output of a model can
be used as an input for another model.
The generated object-relational mapping files
contain the necessary information for representing
objects in relation database tables. For example
names of tables and columns to use, data type of
the fields, where to store properties of the objects
and relations between objects.
In all practicality it is a fully automatically
generated persistence layer for the SEAMLESS
system. Higher layers of the system, and some of
the models themselves, use it to retrieve, work
with and store instances of the concepts as defined
in the ontology.

Figure 4. SEAMLESS integrated framework
architecture.
The objective for the final version of the
SEAMLESS system is to move from off-line use
of the Knowledge Manager to a more integrated
use at runtime, for example to be able to
dynamically add concepts and to help users of
different domains to perform integrated assessment
tasks.

5. CONCLUSION
The challenge in integrated modeling is the
conceptual integration. To achieve this, we need
explicit semantics and a shared conceptualization.
A participatory and collaborative approach is a key
success factor for the creation of a common
ontology for models, indicators and raw data.
The development of the SEAMLESS common
ontology was and still is a big challenge that is
performed by a dedicated taskforce. By putting the
ontology in a central position in the project and the
systems architecture, this shared conceptualization
is the basis for generating (Java) source code for
the object classes representing all the concepts and
representing the objects in relational database
tables.
The use of ontology has proved to be very useful if
not essential both for the technical integration of
knowledge in the SEAMLESS Integrated
Framework and in understanding the meaning of
communicated words of the diversity of people
within the project.
6. ACKNOWLEDGEMENTS
We thank all scientists of the SEAMLESS project
who contributed to the concept of using ontology
and the development of the SEAMLESS common
ontology. This work has been carried out as part of
the SEAMLESS Integrated Project, EU sixth
Framework Programme for Research and
Technological Development, Contract No.
010036-2.
The sixth framework programme (FP6) is a
collection of the actions at EU level to fund and
promote research. The main objective of FP6 is to
contribute to the creation of the European
Research Area (ERA) by improving integration
and co-ordination of research in Europe.
Seamless-IF
components
Models
database


Knowledge base
Model
ontology
Knowledge manager
Indicators
Indicator
ontology
Generated code
(ORM)
Source
data
Hibernate
1964

7. REFERENCES
Cash, D. W., Clark, W. C., Alcock, F., Dickson, N.
M., Eckley, N., Guston, D. H., et al. (2003).
Science and Technology for Sustainable
Development Special Feature: Knowledge
systems for sustainable development.
PNAS, 100(14), 8086-8091.
Farquhar, A., Fikes, R., Pratt, W., & Rice, J.
(1995) Collaborative Ontology
Construction for Information Integration
(No. KSL-95-63): Knowledge Systems
Laboratory, Department of Computer
Science, Stanford University.
Gijsbers, P.J.A., Gregersen, J.B., (2005) OpenMI:
glue for model integration. In: Zerger, A.,
Argent, R.M. (Eds.), MODSIM 2005
International Congress on Modelling and
Simulation. Modelling and Simulation
Society of Australia and New Zealand,
December 2005. pp. 648-654.
Gijsbers, P.J.A., Wien, J.E., Verweij, P., Knapen,
R. (2006) Advances in the OpenMI. In:
Proceedings of 7th International Conference
on Hydroinformatics, HIC 2006. Nice,
France, pp. 72-81.
Griffen, E.M. (1997). A first look at
communication theory. New York:
McGraw-Hill Companies Inc.
Gruber, T. R. (1993). A Translation Approach to
Portable Ontology Specifications.
Knowledge Acquisition, 5, 199-220.
Gruber, T. R. (1995) “Toward Principles for the
Design of Ontologies Used for Knowledge
Sharing”, International Journal of Human
and Computer Studies, 43(5/6), 907–928.
Holsapple, C. W., & Joshi, K. D. (2002). A
collaborative approach to ontology design.
Communications of the ACM, 45(2), 42-47.
IEE, (1990) Institute of Electrical and Electronics
Engineers. IEEE Standard Computer
Dictionary: A Compilation of IEEE
Standard Computer Glossaries. New York,
NY: 1990.
Van Ittersum, M. K., F. Ewert, T. Heckelei, J.
Wery, J. Alkan Olsson, E. Andersen, I.
Bezlepkina, F. Brouwer, M. Donatelli, G.
Flichman, L. Olsson, A. Rizzoli, T. Van
Der Wal, J.-E. Wien and J. Wolf (2007),
Integrated assessment of agricultural
systems- a component based framework for
the European Union (SEAMLESS),
Agricultural Systems In Press.
Neches, Robert, Richard Fikes, Tim Finin, Thomas
Gruber, Ramesh Patil, Ted Senator, and
William R. Swartout. (1991) Enabeling
technology for knowledge sharing. AI
Magazine, Vol.12, No. 3, Fall 1991.
Ogden, C. K. & Richards, I. A. (1923) "The
Meaning of Meaning." 8th Ed. New York,
Harcourt, Brace & World, Inc.
Patel-Schneider, Peter F., Patrick Hayes, Ian
Horrocks. (2004) OWL Web Ontology
Language Semantics and Abstract Syntax.
W3C Recommendation. [Available at:
http://www.w3.org/TR/owl-semantics/ ]
Rizzoli, A., Donatelli, M., Athanasiadis, I., Villa,
F., Muetzelfeldt, R., & Huber, D. (2005).
Semantic links in integrated modelling
frameworks. Paper presented at the
MODSIM 2005 International Congress on
Modeling and Simulation. Melbourne,
Australia
Sølvberg, A. (1998) Data and what they refer to. In
P. P. Chen, editor, Concept Modeling:
Historical Perspectives and Future Trends.
Springer Verlag.
Schoemaker, P. J. H. (1993), Multiple Scenario
Development: Its Conceptual and
Behavioral Foundation, Strategic
Management Journal 14(3), 193.
Scholten, Huub, Ayalew Kassahun, Jens Christian
Refsgaard, Theodore Kargas, Costas
Gavardinas and Adrie J.M. Beulens (2007)
A methodology to support multidisciplinary
model-based water management
Environmental Modelling & Software,
Volume 22, Issue 5, Pages 743-759


1965